CochlearMotion: Head Gesture Recognition Leveraging Ear Canal Deformation Sensing

Abstract

Hands-free interfaces have become increasingly popular due to the growing demands for convenient interaction with mobile and wearable devices. Among all of hands-free interfaces, head gesture interaction has shown great potential in providing alternatives in various real-world scenarios, such as interfaces for people with disabilities and Virtual/Augmented Reality applications. However, existing head gesture recognition systems require either Line-Of-Sight or specialized/customized hardware. Additionally, some approaches could raise potential privacy concerns. In this work, we propose CochlearMotion, a novel in-ear wearable system that achieves head gesture recognition by utilizing off-the-shelf earbuds with a built-in microphone. Specifically, we leverage sonar-like techniques to sense the unique deformation of the ear canal, which closely correlated with each head motion for cross-user head gesture recognition. Our extensive experimental evaluation shows that our system can achieve over 95% recognition accuracy for six typical head gestures and works well in various real-world environments and scenarios.